MisoTTS

Overview

MisoTTS from Miso Labs is an open-weights 8B parameter text-to-speech model, released for developers who need controllable, high-fidelity voice generation.[6] Being open-weights, it can be fine-tuned, embedded in custom inference stacks, or integrated with existing LLM pipelines for voice agents.[6] ML engineers should care because it offers a modern, fully controllable TTS backbone without proprietary constraints.